Due to the explicit 'noinline' GCC-7.3 is not able to optimize away the
argument setup of:
apply_ibt_endbr(__ibt_endbr_seal, __ibt_enbr_seal_end);
even when X86_KERNEL_IBT=n and the function is an empty stub which leads
to link errors due to missing __ibt_endbr_seal* symbols like below:
ld: arch/x86/kernel/alternative.o: in function
`alternative_instructions':
alternative.c:(.init.text+0x15d): undefined reference to
`__ibt_endbr_seal_end'
ld: alternative.c:(.init.text+0x164): undefined reference to
`__ibt_endbr_seal'
Remove explicit 'noinline' to help gcc optimize them away.
[Thanks Peter Zijlstra for tweaking the commit log.]
Signed-off-by: Miaohe Lin <[email protected]>
---
arch/x86/kernel/alternative.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 5cadcea035e0..beaf9fc44e2f 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -624,7 +624,7 @@ void __init_or_module noinline apply_ibt_endbr(s32 *start, s32 *end)
#else
-void __init_or_module noinline apply_ibt_endbr(s32 *start, s32 *end) { }
+void __init_or_module apply_ibt_endbr(s32 *start, s32 *end) { }
#endif /* CONFIG_X86_KERNEL_IBT */
--
2.23.0
On 2022/10/11 19:38, Miaohe Lin wrote:
> Due to the explicit 'noinline' GCC-7.3 is not able to optimize away the
> argument setup of:
>
> apply_ibt_endbr(__ibt_endbr_seal, __ibt_enbr_seal_end);
>
> even when X86_KERNEL_IBT=n and the function is an empty stub which leads
> to link errors due to missing __ibt_endbr_seal* symbols like below:
>
> ld: arch/x86/kernel/alternative.o: in function
> `alternative_instructions':
> alternative.c:(.init.text+0x15d): undefined reference to
> `__ibt_endbr_seal_end'
> ld: alternative.c:(.init.text+0x164): undefined reference to
> `__ibt_endbr_seal'
>
> Remove explicit 'noinline' to help gcc optimize them away.
> [Thanks Peter Zijlstra for tweaking the commit log.]
>
> Signed-off-by: Miaohe Lin <[email protected]>
Add cc Peter Zijlstra.
Change log for v2:
Tweak the commit log per Peter Zijlstra. Thanks!
I should have done this in previous thread. Sorry for make noise.
Thanks,
Miaohe Lin
> ---
> arch/x86/kernel/alternative.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
> index 5cadcea035e0..beaf9fc44e2f 100644
> --- a/arch/x86/kernel/alternative.c
> +++ b/arch/x86/kernel/alternative.c
> @@ -624,7 +624,7 @@ void __init_or_module noinline apply_ibt_endbr(s32 *start, s32 *end)
>
> #else
>
> -void __init_or_module noinline apply_ibt_endbr(s32 *start, s32 *end) { }
> +void __init_or_module apply_ibt_endbr(s32 *start, s32 *end) { }
>
> #endif /* CONFIG_X86_KERNEL_IBT */
>
>
Friendly ping ;)
On 2022/10/11 19:38, Miaohe Lin wrote:
> Due to the explicit 'noinline' GCC-7.3 is not able to optimize away the
> argument setup of:
>
> apply_ibt_endbr(__ibt_endbr_seal, __ibt_enbr_seal_end);
>
> even when X86_KERNEL_IBT=n and the function is an empty stub which leads
> to link errors due to missing __ibt_endbr_seal* symbols like below:
>
> ld: arch/x86/kernel/alternative.o: in function
> `alternative_instructions':
> alternative.c:(.init.text+0x15d): undefined reference to
> `__ibt_endbr_seal_end'
> ld: alternative.c:(.init.text+0x164): undefined reference to
> `__ibt_endbr_seal'
>
> Remove explicit 'noinline' to help gcc optimize them away.
> [Thanks Peter Zijlstra for tweaking the commit log.]
>
> Signed-off-by: Miaohe Lin <[email protected]>
> ---
> arch/x86/kernel/alternative.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
> index 5cadcea035e0..beaf9fc44e2f 100644
> --- a/arch/x86/kernel/alternative.c
> +++ b/arch/x86/kernel/alternative.c
> @@ -624,7 +624,7 @@ void __init_or_module noinline apply_ibt_endbr(s32 *start, s32 *end)
>
> #else
>
> -void __init_or_module noinline apply_ibt_endbr(s32 *start, s32 *end) { }
> +void __init_or_module apply_ibt_endbr(s32 *start, s32 *end) { }
>
> #endif /* CONFIG_X86_KERNEL_IBT */
>
>
On Tue, Oct 11 2022 at 19:38, Miaohe Lin wrote:
> Due to the explicit 'noinline' GCC-7.3 is not able to optimize away the
> argument setup of:
>
> apply_ibt_endbr(__ibt_endbr_seal, __ibt_enbr_seal_end);
>
> even when X86_KERNEL_IBT=n and the function is an empty stub which leads
> to link errors due to missing __ibt_endbr_seal* symbols like below:
>
> ld: arch/x86/kernel/alternative.o: in function
> `alternative_instructions':
> alternative.c:(.init.text+0x15d): undefined reference to
> `__ibt_endbr_seal_end'
> ld: alternative.c:(.init.text+0x164): undefined reference to
> `__ibt_endbr_seal'
This is incomplete because the noinline is only visible within
alternative.c, but there is another caller in module.c
This needs an inline stub in the header to be correct.
Thanks,
tglx
On 2022/12/2 19:59, Thomas Gleixner wrote:
> On Tue, Oct 11 2022 at 19:38, Miaohe Lin wrote:
>> Due to the explicit 'noinline' GCC-7.3 is not able to optimize away the
>> argument setup of:
>>
>> apply_ibt_endbr(__ibt_endbr_seal, __ibt_enbr_seal_end);
>>
>> even when X86_KERNEL_IBT=n and the function is an empty stub which leads
>> to link errors due to missing __ibt_endbr_seal* symbols like below:
>>
>> ld: arch/x86/kernel/alternative.o: in function
>> `alternative_instructions':
>> alternative.c:(.init.text+0x15d): undefined reference to
>> `__ibt_endbr_seal_end'
>> ld: alternative.c:(.init.text+0x164): undefined reference to
>> `__ibt_endbr_seal'
>
> This is incomplete because the noinline is only visible within
> alternative.c, but there is another caller in module.c
>
> This needs an inline stub in the header to be correct.
>
I see. I think I should update my gcc instead :) Thanks for your reply.
Thanks,
Miaohe Lin
The following commit has been merged into the x86/urgent branch of tip:
Commit-ID: 023e59d4cee91dd3ca811eb01db2c5fb9f783dc2
Gitweb: https://git.kernel.org/tip/023e59d4cee91dd3ca811eb01db2c5fb9f783dc2
Author: Miaohe Lin <[email protected]>
AuthorDate: Tue, 11 Oct 2022 19:38:03 +08:00
Committer: Thomas Gleixner <[email protected]>
CommitterDate: Fri, 02 Dec 2022 12:54:43 +01:00
x86/alternative: Remove noinline from __ibt_endbr_seal[_end]() stubs
Due to the explicit 'noinline' GCC-7.3 is not able to optimize away the
argument setup of:
apply_ibt_endbr(__ibt_endbr_seal, __ibt_enbr_seal_end);
even when X86_KERNEL_IBT=n and the function is an empty stub, which leads
to link errors due to missing __ibt_endbr_seal* symbols:
ld: arch/x86/kernel/alternative.o: in function `alternative_instructions':
alternative.c:(.init.text+0x15d): undefined reference to `__ibt_endbr_seal_end'
ld: alternative.c:(.init.text+0x164): undefined reference to `__ibt_endbr_seal'
Remove the explicit 'noinline' to help gcc optimize them away.
Signed-off-by: Miaohe Lin <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
---
arch/x86/kernel/alternative.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 5cadcea..beaf9fc 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -624,7 +624,7 @@ void __init_or_module noinline apply_ibt_endbr(s32 *start, s32 *end)
#else
-void __init_or_module noinline apply_ibt_endbr(s32 *start, s32 *end) { }
+void __init_or_module apply_ibt_endbr(s32 *start, s32 *end) { }
#endif /* CONFIG_X86_KERNEL_IBT */